Breakthrough in Fully Encrypted Computing: 5th-Gen GL FHE Scheme Makes Private AI a Reality

Taipei — At the FHE.org 2026 Conference, South Korean deep-tech startup DESILO, together with Fully Homomorphic Encryption (FHE) inventor Craig Gentry, unveiled the fifth-generation Gentry-Lee (GL) homomorphic encryption scheme. This breakthrough achieves significant improvement in matrix multiplication performance, marking a key milestone in making "Private AI" a practical reality.

Fully homomorphic encryption is a technology that allows computations to be performed directly on encrypted data without first decrypting it. Traditional encryption schemes require data to be decrypted before processing, exposing it to risk during that period. FHE fundamentally solves this problem—users can encrypt their sensitive data and send it to AI models for processing; the service provider completes calculations in an encrypted state and returns results, with the original data remaining encrypted throughout the entire process. This achieves true "data usable but not visible."

Craig Gentry is a pioneer in the field of fully homomorphic encryption. In 2009, he published a groundbreaking paper proving the feasibility of FHE. Since then, he has continued to advance this technology. The GL scheme is his latest research achievement, deeply optimized for the biggest performance bottleneck in current AI applications—matrix multiplication operations.

"The fifth-generation GL scheme represents a major leap in the fully homomorphic encryption field," said DESILO's CTO at the launch. "By redesigning the core algorithms for encrypted operations, we've successfully improved matrix multiplication efficiency to a level usable in actual AI applications. This means enterprises can finally deploy AI models to process sensitive data at scale without sacrificing privacy."

Industry analysts believe GL scheme breakthroughs will first be applied in: Healthcare—hospitals can analyze patient genetic data in encrypted states, AI diagnostic models can provide recommendations without accessing original medical records; Finance—banks can perform risk modeling and fraud analysis while protecting customer privacy; Government—government departments can collaborate across agencies without data leaving borders or being decrypted.

As global data privacy regulations become increasingly stringent, the practical application of FHE technology has become more critical. The EU's GDPR, China's Data Security Law, and other regulations impose strict requirements on data processing—FHE provides technological assurance for compliance.

Industry insiders expect that with the maturation of new-generation technologies like the GL scheme, the private AI market will experience explosive growth over the next two to three years. DESILO is already collaborating with multiple cloud service providers and AI companies to advance commercialization of this technology.

Reference Sources: Manila Times, Investing.com